forkless/strix
1
0
Fork 0
mirror of https://github.com/usestrix/strix.git synced 2026-08-16 01:16:40 +02:00
Code Issues Packages Projects Releases Wiki Activity

docs(prompts,skills): scope cleanup to own files; dedupe JSONL by URL

Browse Source 
Address Greptile review:
- system_prompt: only clean up your own task's files; don't delete
  another agent's files in the shared workspace unless confirmed unused.
- katana.md: extract+dedupe URLs with jq before removing raw .jsonl
  (sort -u on JSONL compares whole records, not URLs).
This commit is contained in:
Ahmed Allam
2026-07-16 04:47:56 -07:00
committed by Ahmed Allam
parent 4b619d57a0
commit 98990bae45
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
strix/agents/prompts/system_prompt.jinja
+1 -1
View File
@@ -244,7 +244,7 @@ DISK & SCRATCH HYGIENE:
- /workspace is a shared, finite disk used by all agents at once — be a considerate tenant - /workspace is a shared, finite disk used by all agents at once — be a considerate tenant
- Prefer bounded recon: scope crawls and scans by depth, duration, and target rather than "collect everything" - Prefer bounded recon: scope crawls and scans by depth, duration, and target rather than "collect everything"
- Send large tool output to /workspace/scratch/, and once you've extracted what you need (e.g. a URL/endpoint list), remove the raw output - Send large tool output to /workspace/scratch/, and once you've extracted what you need (e.g. a URL/endpoint list), remove the raw output
- If disk gets tight or a write fails for space, check what's large under /workspace and clean up before continuing - If disk gets tight or a write fails for space, check what's large under /workspace and clean up files from your own task; leave another agent's files unless you've confirmed they're no longer in use
MANDATORY INITIAL PHASES: MANDATORY INITIAL PHASES:
strix/skills/tooling/katana.md
+1 -1
View File
@@ -63,7 +63,7 @@ Keeping output manageable:
- katana does not cap its own output size, so bound crawls with `-ct` (crawl-duration) and `-d` (depth); on large sites an unbounded `-jsl` / `-kf all` deep crawl can grow very large. - katana does not cap its own output size, so bound crawls with `-ct` (crawl-duration) and `-d` (depth); on large sites an unbounded `-jsl` / `-kf all` deep crawl can grow very large.
- Reserve deep JS crawling (`-jsl`, `-kf all`, higher `-d`) for a specific narrowed target rather than broad scopes. - Reserve deep JS crawling (`-jsl`, `-kf all`, higher `-d`) for a specific narrowed target rather than broad scopes.
- After a crawl, glance at output size (`du -sh <out>`); if it looks outsized for the scope, tighten `-d`/`-ct`/`-ef` or split per host. - After a crawl, glance at output size (`du -sh <out>`); if it looks outsized for the scope, tighten `-d`/`-ct`/`-ef` or split per host.
- Dedupe and reduce before use (`sort -u`), and remove the raw `.jsonl` once you've parsed the URLs you need. - For JSONL output, extract and dedupe URLs before use (`jq -r '.url' <out.jsonl> | sort -u > urls.txt`), then remove the raw `.jsonl` once you've preserved any metadata you still need.
Usage rules: Usage rules:
- Keep `-d`, `-c`, `-p`, and `-rl` explicit for reproducible runs. - Keep `-d`, `-c`, `-p`, and `-rl` explicit for reproducible runs.